【摘要】 地震波场模拟是叠前逆时偏移方法的核心计算单元,因此,边界条件和存储方案的选择是高精度以及高效率逆时偏移算法不可回避的问题.本文就随机边界条件和吸收边界条件在逆时偏移算法中的应用效果以及相应的四种存储策略展开具体分析,给出了不同存储策略下的计算成本和存储量需求.文中优选随机边界条件和吸收边界条件的有效边界存储方案,引入GPU并行加速技术对SEG起伏地表模型进行测试,详细讨论分析了两种方法的计算效率和计算精度.测试表明,随机边界存储的计算成本要低于有效边界存储的计算成本,而受边界散射的影响,随机边界逆时偏移在浅层伴有噪音干扰.针对具体数据情况,合理地运用GPU加速技术、选择有效的边界条件以及存储方法,是改善逆时偏移成像精度和效率的有效途径.更多还原

【Abstract】 Seismic wave field modeling is the computing core of pre-stack reverse time migration.Therefore,for improving the accuracy and computational efficiency of reverse time migration,effective boundary condition and storage strategies are the key factors.In this paper,we analyzed the application effect by reverse time migration and four kinds of storage strategies with random and absorbing boundary conditions,summarized the calculation cost and storage requirement for different boundary and storage strategies.We have tested the reverse time migration algorithm on SEG rugged topography model with random boundary condition and absorbing boundary condition with its efficient boundary storage scheme chosen,combined with GPU(Graphic Processing Unit) acceleration technology.The result shows that the computational cost of random boundary storage is lower than which of efficient boundary strategy,but affected by the boundary scattering,the reverse time migration profile with random boundary would get noise interference in shallow layers.The key scheme to improve the accuracy and computational efficiency of reverse time migration is choosing compatible boundary condition,storage strategy,and GPU acceleration technology according to the specific data.更多还原